Research map / Probabilistic and scientific modeling / Causal discovery
Structural Causal 3D Reconstruction
Weiyang Liu, Zhen Liu, Liam Paull, Adrian Weller, Bernhard Schölkopf
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add BO-searched dense causal ordering among latent factors | Independent SCR · BFM depth (CUDA-10) | MAD (deg.): 17.36 -> 14.88 | Improved |
| add BO-searched dense causal ordering | Independent SCR · BFM depth (CUDA-10) | SIDE x1e-2: 0.895 -> 0.830 | Improved |
| use random dense ordering instead of searched ordering (ablation) | Dense SCR · BFM depth (CUDA-10) | SIDE x1e-2: 0.830 -> 1.000 | Hurt |
| learn DAG over factors with dense-ordering mask (Eq. 3) | Dense SCR · BFM depth (CUDA-10) | SIDE x1e-2: 0.830 -> 0.820 | Improved |
| learn input-dependent DAG with cosine | Generic SCR · BFM depth (CUDA-10) | SIDE x1e-2: 0.859 -> 0.815 | Improved |
| use DAVL ordering instead of DVAL for dense mask (ablation) | Generic SCR · BFM MAD | MAD (deg.): 14.85 -> 14.79 | Improved |
| use CUDA-9 instead of CUDA-10 | Dense SCR (BO) · BFM depth | SIDE x1e-2: 0.830 -> 0.693 | Improved |
| add dense SCR | Independent SCR · Face pose preservation | angle difference: 12.3 -> 11.0 | Improved |
| add dynamic SCR | Independent SCR · Face pose preservation | angle difference: 12.3 -> 10.5 | Improved |
| add generic SCR | Independent SCR · BFM depth (CUDA-9) | MAD (deg.): 16.26 -> 13.22 | Improved |
Closest papers by meaning
- Learning latent causal graphs via mixture oracles (NeurIPS 2021 · similarity 0.87)
- Unsupervised Causal Generative Understanding of Images (NeurIPS 2022 · similarity 0.87)
- Multi-View Causal Representation Learning with Partial Observability (ICLR 2024 · similarity 0.86)
- Decompositional Neural Scene Reconstruction with Generative Diffusion Prior (CVPR 2025 · similarity 0.85)
- Sparse-view Pose Estimation and Reconstruction via Analysis by Generative Synthesis (NeurIPS 2024 · similarity 0.85)
- The More You See in 2D the More You Perceive in 3D (CVPR 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
